REM sleep is almost never normally entered from wakefulness except in infants. It is preceded by NREM sleep, usually stages 1 and 2. It can therefore be regarded as a type of arousal from NREM sleep _[12]. There are two drives which appear to determine whether or not REM sleep occurs.

Homeostatic drive
This increases with the duration since the previous REM sleep episode in a similar manner to the NREM sleep homeostatic drive, and decreases during REM sleep _[13].

It increases throughout episodes of NREM sleep, but wakefulness appears to be a substitute for REM sleep in this respect and the REM sleep homeostatic drive only increases slightly in wakefulness. As a result there is little pressure to enter REM sleep at the onset of sleep, but this builds up increasingly during the night after episodes of NREM sleep. During wakefulness, although there may have been a considerable interval since the last episode of REM sleep, the tendency to manifest features of REM sleep is limited to a 90-min cycle of transient feelings of drowsiness or a tendency to daydream.

Circadian rhythm
The tendency to enter REM sleep peaks at around 9.00 am independently of the duration of previous REM or NREM sleep. This appears to be under circadian control and may be related to the increase in catabolic hormone secretion, such as cortisol, and the reduction in production of anabolic hormones.

As a result of these two drives, REM sleep appears at approximately 90-min intervals during the night.

This ultradian rhythm may simply be a reflection of the two separate drives to enter REM sleep and their balance with NREM sleep and wakefulness. No specific anatomical location for this drive has been found. The increased duration of REM sleep later in the night is probably due to the combination of an increase in homeostatic and circadian drives to enter REM sleep and the reduced homeostatic NREM sleep drive towards the end of the night.

References